Circuit mechanisms for learning in the rodent Prefrontal cortex and their dysfunction in Schizophrenia

(1)

1

Circuit mechanisms for learning in the rodent Prefrontal cortex and their

dysfunction in Schizophrenia

Inauguraldissertation

zur

Erlangung der Würde eines Doktors der Philosophie vorgelegt der

Philosophisch-Naturwissenschaftlichen Fakultät der Universität Basel

von

Arghya Mukherjee

von Asansol, Indien

Originaldokument gespeichert auf dem Dokumentenserver der Universität Basel

edoc.unibas.ch Basel, 2018

(2)

2

Genehmigt von der Philosophisch-Naturwissenschaftlichen Fakultät auf Antrag von

Prof. Dr. Pico Caroni

(Fakultatsverantwortlicher und Dissertationsleiter)

Prof. Dr. Andreas Luethi (Koreferent)

Basel, 19.09.17

Prof. Dr. Martin Spiess (Dekan)

(3)

3

Preface

Humans and other higher mammals do more than just learn associations between sensory cues and outcomes. We engage in complex top down control of behavior where our internal states and intentions drive us to choose actions that are distant from our goals. Our ability to make such choices not only helps in predicting the outcome of current behavioral contexts and direct our responses as per our needs, but also shapes our future behavioral strategies. One of the most enduring mysteries of the brain is this process of decision-making. What neural mechanisms helps us choose a course of action between several different possibilities? How do we decide to stick with a response and how do we decide to switch to an alternative response?

Ultimately, this seemingly willful but complex behavior emerges from interactions between neurons.

One important feature of the neural mechanisms behind decision-making is that they are shaped by experience. Thus, they depend on areas of the brain, which can learn the ‘rules of the game’ – what outcomes are desirable and what responses can be directed towards achieving these outcomes? The structure, most frequently, implicated in such top down control is the Prefrontal cortex (PFC). Indeed the complexity of the PFC structure and functions is greatest in humans and much simpler in mammals such as rodents, which parallels the ability of higher order primates to make complex decisions. Moreover, the effects of PFC dysfunction in humans are most apparent where cognitive control is needed.

In this thesis, I explore the mouse prefrontal cortex, primarily the Prelimbic and Infralimbic cortices, which are known to play antagonistic roles in goal-oriented behavior. Using recently developed chemogenetic tools, I will show how these two areas are connected with each other and the functional significance of these connections in learning competing associations. In a second part, I will show how local neuronal networks within the PFC and the hippocampus are affected in a mouse model of schizophrenia and propose a clinically relevant therapeutic strategy for long- term amelioration of the cognitive symptoms of this disorder.

(6)

6

Rationale and Goals

Flexible behavior, as shown by most mammals, requires continuous decision making where appropriate actions must be chosen from an array of available actions based on our current goals and prior experience. The medial prefrontal cortex (mPFC) is essential for selecting such appropriate actions and inhibiting inappropriate ones. The prefrontal cortex is not a homogenous structure but rather an agglomeration of sub- areas, which sub serve different functions. For example, the anterior cingulate is required for effort-based decision-making (Walton et al., 2003; Rushworth et al., 2004) while the orbitofrontal cortex is essential for value based decision-making (Niv and Schoenbaum, 2008). However, the outcome of a decision making process is selection of a singular behavioral action or learning a new association. Hence, it would be reasonable to hypothesize that this selection would be a product of the combined output of the various prefrontal areas and the interactions among them. Thus, to understand the neurobiological substrates of decision making one needs to explore the prefrontal cortex at two different levels: 1. The internal microcircuit and neuronal networks within individual prefrontal areas, and 2. Functional interactions among the prefrontal areas. The broad goal of my thesis was to use both of these approaches to study the prefrontal cortex of a well-established model organism (mouse) which has a relatively simple behavioral repertoire yet is evolutionarily complex enough to generalize my findings to higher order animals.

First, I focused my attention on the Prelimbic (PreL) and Infralimbic (IL) regions of the mouse medial prefrontal cortex (mPFC). These two areas have been studied most extensively among the rodent prefrontal areas. In several behavioral domains, the PreL and IL exert distinct and opposing, influences over behavior; in a PreL-Go/IL- NoGo manner. The most common examples of this complementary function are the expression and extinction of conditioned fear responses or drug seeking behavior (Peters et al., 2009). Furthermore, neuronal tuning studies have shown that the PreL neurons are tuned to the representation of goals in goal directed learning while the IL neurons appear to tune to alternative choices. I investigated how the PreL and IL cortices interact among each other to influence learning and selection of behavioral strategies. Such, interactions between IL and PreL or other prefrontal areas have not

(7)

7

been studied in detail in the past with one notable exception. Research done by Ji and Neugebauer (2012) have shown that optogenetic activation of IL inhibits PreL pyramidal cells in vivo, implying an existence of feed-forward inhibition from the IL to PreL. I carried out selective chemogenetic silencing of PreL or IL during different sub phases of the Intra-dimension/ extra-dimension set shifting task (IEST) or trace learning and extinction to evaluate their individual contributions. My findings suggest that PreL promotes application of behavioral strategies or new learning corresponding to previously learnt associations while IL is required to learn alternative associations across different learning paradigms. Next, using viral mediated tracing techniques I show the existence of reciprocal layer5/6 derived IL

↔

PreL projections. Using selective unidirectional silencing/activation of these projections, I have shown that the ILàPreL and PreLàIL projections are required at different phases of learning.

Unidirectional ILàPreL projections are specifically required during IL mediated alternative learning (eg: extinction) and bi-directional PreL

↔

IL projections are required +12-14h post learning to setup the role of IL in subsequent learning of alternative choices.

Prefrontal cortex dysfunction has been identified as a key neurobiological correlate of cognitive deficits associated with many neuropsychiatric disorders like Schizophrenia, Attention deficit/Hyperactivity disorder etc. Exploring the dysfunction of defined prefrontal neuronal networks and circuits in rodent models of neuropsychiatric disorders can be also be a rewarding approach towards understanding decision making. In the second part of the thesis, I explored the dysfunction in the Parvalbumin (PV) interneuron network in a mouse model of Schizophrenia.

Parvalbumin interneurons have been shown to synchronize network activity, supporting different types of neuronal network oscillations, such as gamma and theta oscillation, ripple and spindle activity (Amilhon et al., 2015; Stark et al., 2014; Lapray et al., 2012). Thereby, they play a significant role in the formation and consolidation of memories to support learning and behavior (Karunakaran et al., 2016; Donato et al., 2013). Finally, dysfunction of the Parvalbumin interneuron system, in the prefrontal cortex of human schizophrenia patients, has emerged as a core substrate underlying the cognitive deficits in the disease (Lewis, 2014). Thus, studying the dysfunction of

(8)

8

the PV network in Schizophrenia not only provides a way to understand its role in prefrontal function but also raise the possibility of developing therapeutic strategies to ameliorate the associated cognitive deficits. I first showed that the PV network in LgDel+/- animals fail to mature with respect to those of their wild type counterparts and remain stuck in an immature state, which is also associated with altered neural synchrony in the gamma band and behavioral deficits. I further show that stimulation of the PreL PV neuron network within a specific window of treatment during early adulthood can rescue the dysfunctional PV network synchrony as well as behavioral deficits. In recent years, interactions between the hippocampus and prefrontal cortex (PFC) have emerged as key players in various cognitive and behavioral domains (Harris and Gordon, 2015). Disruptions in hippocampal-prefrontal interactions have also been observed in psychiatric disease, most notably schizophrenia (Godsil et al., 2013). I found that long-term rescue of the PreL PV state and associated behavioral deficits in LgDel+/- mice can also be mediated through direct stimulation of the ventral hippocampal (vH) PV network. However if the rescue is targeted to PreL while preventing it in vH or vice versa, it fails to mediate any behavioral rescue in LgDel+/- mice. Thus suggesting that long-term rescue of the PV pathology and cognitive deficits in LgDel+/- animals requires a rescue of the entire hippocampal-prefrontal axis.

(9)

9

1. Introduction

(10)

10 1.1. The Prefrontal cortex

Located in the anterior end of the mammalian brain the Prefrontal cortex is a collection of anatomically and functionally distinct brain regions, which play different roles in cognitive control. The prefrontal areas have a unique connectivity pattern with almost all neocortical and some subcortical structures (Bedwell et al., 2014, Kondo and Witter, 2014, Hoover and Vertes, 2011, Vertes, 2004). This provides a perfect framework to integrate a diverse range of information needed for top down control of behavior. Such behavior is largely driven by rule learning, where new information is incorporated and processed in the context of previous experience to result in ‘rules’ for guiding future behavior. Thus, traditionally, the PFC has been studied in the context of decision- making processes like working memory, error detection (Holroyd et. al., 2002), reinforced learning and its extinction (Rushworth et al., 2011)

The PFC has also been in recent focus due to a growing recognition that dysfunction in its development and circuitry may underlie cognitive symptoms associated with disorders such as Schizophrenia (Volk and Lewis, 2010), Bipolar disorder and Attention-deficit/Hyperactivity disorder (Schubert et al., 2015). Although the extent to which cognitive processes are functionally homologous across humans, non-human primates and rodents remain controversial, rodent models of such disorders have proven to be an invaluable tool for studying these processes. Taking a reductionist approach, neurobiologists working with rodents have been able to allocate behavioral endo-phenotypes to anatomical subdivisions within the prefrontal cortex with increasing specificity. This process is aided by recently developed microscopic techniques, which allow us to observe neuronal activity in prefrontal cortical areas at single neuron resolution in freely behaving mice/rats. Moreover, chemogenetic and optogenetic manipulations of well-defined assemblies of neurons with high temporal and spatial resolution offers a possibility to map the dysfunctions at the level of microcircuits, cell assemblies and molecular processes. This in turn offers newer targets for drug development and novel therapeutic strategies. In the following sections, I will describe the structure and function of different rodent prefrontal areas.

I will also describe mnemonic functions of the medial PFC in a broader context of its role in goal directed behavior and decision-making. Finally, I will explore the

(11)

11

dysfunction of the medial PFC (mPFC) in the context of psychiatric and cognitive disorders.

1.2. Structural organization of the rodent Prefrontal Cortex

Comparison of neuroanatomy across multiple species show that the evolution of the Prefrontal cortex parallels the evolution of cognition. Thus, higher order mammals such as humans and non-human primates have a highly elaborated PFC. The areas that comprise the monkey PFC are often grouped into anatomically different subfields, namely, the orbital and medial (Brodmann area 10, 11, 13, 14), the lateral (Brodmann areas 46, 12, 45), and the mid-dorsal (Brodmann area 9) regions. The earliest descriptions of the prefrontal cortex were based on anatomical criterion like cortical projection areas of the mediodorsal thalamic nuclei or having a granular layer IV and a location rostral to the agranular premotor areas (Barbas and Pandya, 1989). Yet such cytoarchitectonic criterions for defining the prefrontal cortex are not enough when comparing distantly related species such as monkeys and mice. For example, unlike in primates, the rodent PFC is agranular and lacks a discernable layer IV. On the other hand, primate and rodent PFC share similar patterns of projections from the mediodorsal thalamus. Thus, the rodent PFC is defined based on multiple levels of comparison with the primate PFC, beyond just cytoarchitectonics (Ongur and Price, 2000).

The criteria taken into consideration for defining a rodent PFC homologous to the primate PFC are: 1) similarities in input-output connectivity, 2) functional homologies at the level of single cells, network activity and behavioral roles, 3) distribution of receptors and neuro-modulatory projections and 4) development and maturation criterion (Uylings et al., 2003). Based on these criterions the rodent prefrontal cortex comprises of a medial prefrontal cortex (mPFC), a lateral PFC (lPFC) and a ventral region (vPFC). The medial PFC can be further subdivided into a dorsal part (dmPFC) consisting of the Anterior Cingulate cortex (ACC) and the Prelimbic cortex (PreL), and a ventral part (vmPFC) consisting of the Infralimbic cortex (IL) and medial Orbital cortex (MO). The lateral subdivision of the rodent PFC includes the dorsal and ventral Insular areas (AID, AIV) and the lateral Orbital cortex (LO). Lastly, the ventral PFC encompasses the ventral Orbital and the ventrolateral Orbital cortex (VLO). However,

(12)

12

most functional and anatomical studies of the rodent PFC have focused on a broader subdivision of the PFC, namely mPFC (ACC, PreL, IL) and the Orbitofrontal PFC (MO, LO, VO, VLO). Hence, in the subsequent sections I will focus my discussion on these regions.

1.2.1 Anterior Cingulate cortex

The rat anterior cingulate cortex (ACC) is an integral part of the prefrontal cortex and lies on its medial surface rostral and caudal to the genu of the corpus callosum. It consists of Layers I, II/III, V and VI and lacks a layer IV (Devinsky et al., 1995). Based on neuroanatomy the ACC can be further subdivided into a dorsal (ACCd) and a ventral (ACCv) subpart. In ACCd the cells are arranged in a columnar fashion in comparison to the ACCv (Van de Werd, 2010). Furthermore, in LVI the cells in ACCd are columnar in arrangement while those of ACCv are arranged in horizontal rows.

Like other cortical areas, Layer I of ACC mainly contains interneurons and fibres projecting from other parts of the brain. Some of these incoming fibers also form connections in the deeper layers of the ACC. Layers II-III mainly consists of pyramidal neurons while layers V-VI contain both pyramidal neurons and interneurons. The dendrites of deep-layer pyramidal cells spread into the superficial layers of the ACC (Wu et. al., 2009), and so it is likely that neurons between different layers of ACC also form synaptic connections. However, such intra layers connections and their functional relevance in the ACC has remained largely unexplored.

The projections into the ACC can be broadly classified into three systems. The first set of projections originate from the medial thalamus, which in turn receive projections from the spinothalamic tract (Yang et al., 2006; Shyu et al., 2009). These thalamic projections were recently shown to directly excite Parvalbumin interneurons in the ACC to mediate feed forward inhibition onto Layer II-III neurons (Delevich et. al, 2015).

A second set of inputs originate from the Amygdala, specifically from the central nucleus while a final set of inputs originate from other sensory cortical areas like the somatosensory cortex (S1) and the Insular cortex (Han et al., 2015). In turn, the ACC sends out projection to subcortical structures like the hypothalamus, periacquedectal grey and dorsal horn of the spinal cord via layer V pyramidal neurons (Bragin et al., 1984). ACC neurons also send reciprocal projections to the amygdala, specifically the

(13)

13

BLA, and to the main norepinephrine producing brain sites – locus coeruleus (LeDoux, 2000; Tovote et al., 2015). The ACC also has bi-directional connectivity to the neighboring mPFC areas such as PreL and IL (Medalla and Barbas, 2012) as well as the Retrosplenial cortex (RSC). Importantly the ACC is part of the Papez circuit where it receives inputs from the anterior thalamic tracts and sends outputs to the hippocampal formation via the entorhinal cortex (Jankowski et al., 2013). Thus, the ACC has a significant role to play in the control of emotional expressions. Recent studies have also shown a direct projection from the ACC to the dorsal hippocampus (dH) which is required for the expression of learnt fear (Rajasethupathy et. al., 2015).

The overall connectivity of the ACC show an anterior posterior gradient where connections to and from the limbic and emotion processing regions are more segregated to the anterior side while the posterior side mostly communicates with dorsal and lateral frontal areas and is thought to be involved in top-down regulation (Bliss et al, 2016; Etkin et al, 2011).

Figure 1.1: Input/output connectivity of ACC: The ACC receives various inputs from nociceptive centres like the thalamus as well as from emotive centres such as the amygdala. In turn the ACC sends direct projections to the dorsal horn of the spinal cord. This loop has been predicted to be important for central sensitization. ATN: Anterior thalamic nuclei; LC: Locus coeruleus; dH: dorsal Hippocampus;

EntC: Entorhinal cortex; S1: primary Somatosensory area.

1.2.2. Orbitofrontal cortex

The Orbitofrontal cortex is so called due to its location around the eye socket in primates. In rodents, the OFC is determined as those areas of the frontal lobe that have strong anatomical and functional parallels with the primate OFC. Thus, the rodent

(14)

14

OFC encompasses the dorsal bank of the rhinal sulcus including the lateral orbital regions – MO, LO, VO and VLO (Groenewegen 1988; Ray and Price, 1992). Like other PFC areas the OFC is also comprised of Layers I, II, III, V and VI and a missing LIV. The border between layers II and III are indistinguishable in VO and VLO while MO and LO have a clear border between these two layers (Van de Werd, 2010). The various OFC areas can also be distinguished based on the distribution of Parvalbumin positive interneurons (PV), which are the primary source of perisomatic inhibition in the cortex. The LO and VLO have a high density of these interneurons spread across all layers while the MO has a sparse distribution of these interneurons. The OFC can also be distinguished by the absence of dopaminergic afferents. Specifically, the MO, VLO and LO are largely bereft of dopaminergic projections while the VO has dopaminergic fibers in its caudal part. Finally, like the ACC, intra layer connectivity and its functional significance has not been explored in the OFC (Van de Werd, 2010).

With respect to its input/output connectivity, the OFC is positioned at the intersection of multimodal sensory networks and circuits involved in emotion and memory. A chief characteristic of the orbital network is that it receives inputs from all sensory cortical systems, including olfaction, taste/visceral afferents, vision, and somatic sensation (Rolls, 1996; Groenewegen and Uylings, 2000; Ongur and Price, 2000; Hoover and Vertes, 2007). However, although the OFC has extensive sensory inputs, it only weakly connects with the motor system. It also has strong connections with the limbic system, including the hypothalamus, amygdala, and hippocampus and nucleus accumbens (Rempel-Cowler, 2007b; Floyd et al., 2001). Of the limbic areas, the OFC connects strongest with the amygdala. All parts of the orbitofrontal cortex receive input from the amygdala, primarily originating from the basolateral, basomedial, and lateral nuclei. The medially situated VO and VLO areas issue projections to the anterior portion of the amygdala, with sparse termination distributed across the basomedial nucleus, medial nucleus, central nucleus and intercalated nuclei. Notably, the VO and VLO sends projections to autonomic output areas, including the medial and central nuclei of the amygdala, and issue only weak projections to the basolateral amygdala.

Taken together these OFC projections to sensory emotion and reward-related regions suggest that the OFC is likely to be important in encoding the value of incoming stimuli.

(Hoover and Vertes, 2011; Rempel-Cowler, 2007a)

(15)

15

Figure 1.2: Input/output connectivity of OFC: The OFC receives inputs from all sensory modalities.

It also sends outputs to higher order cortical areas like PreL, ACC and RSC as well as to subcortical areas, which regulate actions, such as accumbens, lateral hypothalamus, PAG etc. Of all cortical areas the OFC has the densest reciprocal connections across the amygdalar complex. This pattern of connectivity puts the OFC in a unique position to integrate sensorimotor inputs with action selection based on emotive states. TEa: Temporal association area; RSC: Retrosplenial cortex; MD: Medio- dorsal thalamus; LH: Lateral hypothalamus.

1.2.3 Ventromedial PFC: Prelimbic and Infralimbic cortex

Lying next to each other in the dorso-ventral axis, the Prelimbic (PreL) and Infralimbic (IL) cortices form the ventral subdivision (vmPFC) of the medial PFC of rodents. The rodent vmPFC exhibits laminar organization with deep and superficial layers and lack a discernable layer IV like the rest of the rodent PFC (Caviness, 1975; Yang et al., 1996; Uylings et al., 2003; Van de Werd et al., 2010), Anatomically the PreL cortex is easily distinguished from the IL by the structure of cortical layers. For example, layer V of PL is less well organized compared to more dorsal regions (i.e., ACC), whereas layer VI cells are arranged in a horizontal fashion. Moreover, in the PreL layers II, III and V are clearly distinguishable, where a densely packed layer II is separated from layer V by a sparsely packed layer III. On the other hand, a unique feature of IL is that cells of layer II spread far into layer I, while in PreL only few cells of layer II are seen in layer I. Therefore, layer II appears wider in IL than in PreL. In general, IL layers II-

(16)

16

VI have a relatively homogenous layout in terms of cell size and density, with smaller cell bodies compared to PreL (Uylings and van Eden, 1990; van de Werd et al., 2010).

Another discerning feature of the PreL and IL is a relative sparse distribution of Parvalbumin interneurons compared to other PFC areas. Moreover, these interneurons are restricted mostly to layer V and VI. A further distinction between PreL and IL can be made with respect to dopaminergic projections, where the PreL receives substantially more DA afferents than the IL (van de Werd et al., 2010).

Apart from the thalamus and parts of the cortex, PreL and IL areas have functionally and anatomically segregated outputs. One such area of differential projection is the Nucleus accumbens (NAcc), an area involved in reward signal processing. PreL fibers distribute extensively throughout the core and shell regions of NAcc. By contrast, IL fibers project selectively to the medial shell of the accumbens (Vertes, 2004). IL and PreL also project very differently to the amygdala. IL fibers distribute widely throughout the anterior part of the amygdala, mainly to rostral and medial Amygdala (MEA), the capsular and medial subdivisions of the central Amygdala (CEA), and to the basomedial nucleus. By contrast, the PreL fibers selectively target the central nucleus (capsular portion) and the BLA. IL and PreL also project differentially to the hypothalamus. IL projects significantly to the dorsomedial hypothalamic nucleus/area, the lateral hypothalamus and supramammillary nuclei (McDonald et al., 1996; Vertes, 2004). On the other hand, PreL fibers mainly pass through the hypothalamus and terminate in the brain stem. The IL and PreL also send differential projections to the brain stem. The IL efferents mainly target SN, a primary site for dopamine production and the ventrolateral regions of the pontomesencephalic PAG. By contrast, PreL efferents project mainly to the VTA and to a lesser extent to the SNc as well as to the ventrolateral pontine PAG (Floyd et al., 2000; Vianna and Brandão, 2003; Hoover and Vertes,2007). Within the cortices the PreL and IL have primarily ipsilateral and largely overlapping connections. For example, IL and the PreL both project to the ACC, entorhinal cortex, piriform cortex and insular cortex (Vertes, 2004). The input connectivity into the mPFC is less segregated between the PreL and IL with notable exceptions. A recent study (Senn et al., 2014) showed that the IL and PreL receive reciprocal connections from distinct, non-overlapping population of neurons in the basal amygdala. In addition, BLA projections also synapse a small percentage of PV interneurons in the PreL and IL (Gabbott et al., 2006). Thus, BLA projections can

(17)

17

functionally modulate mPFC output via feed-forward inhibitory mechanisms.

Furthermore, a population of ventral CA1 neurons synapses onto layers I and V of the IL and these same hippocampal neurons also synapse on entorhinal neurons, which may be important for integrating contextual and spatial information (Swanson, 1981).

With the above exceptions, the PreL and IL receive inputs from most sensory association cortices in the temporal and parietal lobes as well as the other prefrontal areas like OFC and ACC. This suggests that the PreL and IL might integrate incoming information from multiple sources to drive appropriate behavioral responses. However, the cortico-cortical interactions between IL and PreL have not been studied in detail.

A study in slice cultures has shown that the IL has higher frequency local field potentials (LFP) than PreL, and these differ when the two regions are disconnected—

implying some level of functional connectivity between them (van Aerde et al., 2008).

In addition, optogenetic activation of IL inhibits PreL pyramidal cells in vivo (Ji and Neugebauer, 2012) – implying an existence of feed-forward inhibition from the IL to PreL.

Figure 1.3: Input/output connectivity of PreL and IL: The PreL and IL are reciprocally connected to all cortical and sub-cortical areas involved in goal-directed behavior. They also receive direct inputs from the ventral hippocampus. ACC: Anterior cingulate cortex; OFC: Orbitofrontal cortex; NAc: Nucleus Accumbens; vH: ventral Hippocampus; VTA: Ventral tegmental area; PAG: Periacquadectal grey.

(18)

18

1.2.4 The cortico-basal ganglia-thalamo-cortical loop

Other than the region-specific connections of the prefrontal areas, the PFC is part of an important neuronal network in the brain whose dysfunction has been implicated in a number of disorders such as Schizophrenia: the cortico-basal ganglia-thalamo- cortical loop. Within the prefrontal cortex three such loops have been described, each originating from a different prefrontal area. Cortico-striatal terminals are primarily from the LV and LIII pyramidal neurons and are distributed in patches in the basal ganglia.

Unlike the temporal and posterior regions of the cortex, the PFC has highly organized connections with the basal ganglia via the striato-palladial and striato-nigral circuits.

Projections from PFC and ACC cortical areas terminate primarily in the rostral striatum, including the ventral striatum (VS), caudate, and putamen. The vmPFC projection field to the striatum is very limited and is concentrated within the shell and a narrow column along the medial border of the caudate nucleus adjacent to the ventricle (Bonelli and Cummings, 2007). In contrast to the mPFC, the OFC projects to the central and lateral parts of the VS and extend more centrally than the mPFC projections. OFC projections also continue dorsally along the medial regions of the caudate nucleus and ventromedial putamen. However, their location is lateral to the vmPFC projections. Within the OFC projections, there also is some topography such that medial OFC regions terminate medial to those from more lateral OFC areas. The projections from ACC to the striatum are extensive and stretch from its rostral end to the anterior commissure. Terminals are located in both the central caudate nucleus and putamen. Overall, the ACC fibers terminate somewhat lateral to those from the OFC. Thus, the OFC terminal fields are positioned between the vmPFC and ACC, reflecting a topographic arrangement of cortico-striatal projections originating from the PFC. In turn these striatal areas projects to globus pallidus interna (GPi), globus pallidus externa (GPe), and caudolateral SN (Bronstein et al., 2001; Houk, 2001). The globus pallidus (GP) connects to the ventrolateral, ventral anterior, and centromedian nuclei of the thalamus, whose outputs are to supplementary motor area, premotor cortex, and motor cortex. Finally, the thalamic nuclei have reciprocal connections to the putamen and to the frontal cortex thus completing the circuit. Taken together the components of the prefrontal cortex that mediate behaviors are reflected in the

(19)

19

organization and connections between areas of PFC and in their projections to the striatum and thalamus (Bonelli and Cummings, 2007; Haber, 2016).

1.3. Functional roles of the rodent PFC

1.3.1 Anterior Cingulate Cortex

The Anterior cingulate cortex has a role in a variety of cognitive functions. Recent studies have concluded that the ACC plays an important role in mediating instrumental behaviors that require discrimination of multiple relatively similar stimuli. In particular, the ACC could be a structure enhancing stimulus discrimination, if stimuli share common elements, i.e., are in the same sensory modality and are similar (Cardinal et.

al., 2003). Additionally the ACC has also been implicated in effort-based decision- making, where animals prefer high-effort actions, which lead to a larger reward than an alternative low effort action, which leads to a smaller reward while animals with an ACC lesion preferred the low effort reward. Although when the same degree of effort was required to obtain a large or small reward ACC lesioned animals showed preference for the larger reward like sham controls (Walton et al., 2003). This shows that the ACC is important in decision-making processes where an association has to be made between the effort required to obtain a reward and the magnitude of the reward itself (Rushworth et al., 2004).

The ACC has also been the focus of recent studies to understand how nociceptive signals are used to accomplish behavioral goals such as avoiding noxious stimuli and in turn to better understand the cognitive effects of chronic and acute pain. Specifically the anatomical connection of the ACC with areas involved in emotional responses like the amygdala suggest a role for the ACC in processing anxiety and fear in relation to painful stimuli or experiences (Gabbot et al., 2005; Cassel and Wright, 1986;

Buchanan et al., 1994). In all animals, pain can result in expression of conditioned behavior triggered by the context in which the painful event occurred to prevent aggravating the pain. Using classical fear conditioning paradigms recent studies have shown that activation of ACC neurons results in freezing behavior and a conditioned fear memory for the place where the activation occurred (Tang et al., 2005).

Conversely, bilateral inactivation of the ACC inhibits formation of conditioned fear

(20)

20

memory induced by foot-shocks. In rodents trace fear conditioning induces AMPA receptor mediated responses in the ACC and rodent models of chronic pain show consistently impaired trace fear memory as well as increased anxiety like behavior.

Other experiments have shown that high levels of baseline anxiety in rodents contribute to an increase in acute visceral pain responses (Robbins et al., 2007). A recent study also found that selective optogenetic activation of pyramidal neurons of the ACC in mice induced anxiety- and depression-like behavior (Barthas et al., 2015).

These findings, taken together, support the idea that activity in ACC neurons influences anxiety and fear behavior in response to nociceptive events. (Bliss et al., 2016)

1.3.2 Orbitofrontal cortex

Recent studies in rodents and primates indicate that the OFC is crucial for signaling information about expected outcomes and for using these signals to guide flexible behaviour i.e. value based decision-making (Gallagher et al., 1999; Pickens et al., 2003, 2005). Specifically, OFC signals predict characteristics, such as sensory properties (size, shape, texture and flavor) and unique value of specific outcomes that an animal expects given particular circumstances and cues in the environment.

Evidence for such a role comes from behavioral as well as neuronal recording studies.

Neuronal activity studies show that the OFC neurons demonstrate anticipatory firing, which are especially strong before the animal receives a reward or a punishment. For example, in discrimination learning, neurons in the rat OFC demonstrate selective firing in anticipation of sucrose (reward) or quinine (punishment) (Schoenbaum et al., 1998). Initially these neurons fire to one of the two outcomes, with further training they fire in anticipation of the outcome and finally fire in response to the cues predicting the outcome. However, unlike reward-responsive neurons, which also show a transfer of activity from outcomes to cues with training, the OFC neurons do not modulate their firing with respect to unexpected changes in the reward (omission or unexpected delivery) (Schoenbaum et al., 2003; Takahashi et al., 2009). Although such anticipatory firing has been seen in other areas, it first appears in the OFC and persists across all trial events, rather than being triggered by them. Thus suggesting that signaling expected outcomes, be they rewarding or punitive, is a major function of

(21)

21

OFC neurons. Such a hypothesis for OFC function is also backed up by behavioral studies.

Figure 1.4: OFC neurons fire to expected outcomes across all phases of learning: The OFC neurons initially fire after reinforcement in the early phases of a task and with learning, they also fire to the cue itself. These OFC neurons do not stop firing in response to the reward even in later phases of learning, unlike reward responsive VTA neurons fire more strongly in response to unexpected rewards and decrease firing when an expected reward is not delivered. (Schoenbaum et. al., 2009)

A conclusive example of the role of the OFC in signaling expected outcomes is seen in Pavlovian reinforcer devaluation experiments. Here, an animal, which has been trained to associate a cue with a particular reward, is exposed to a protocol where the value of the reward is reduced by pairing it with illness or by feeding it to satiety.

Subsequently the animals’ ability to use that new value to guide its learnt response is assessed by presenting the cue alone. Animals normally show a reduced response to the predictive cue, reflecting their ability to access current lower value of the reward.

However, rats with OFC lesions fail to show this effect of devaluation (Machado and Bachevalier, 2007). The role of OFC is fundamentally different from amygdala, which also shows anticipatory signaling but is only necessary during earlier phases of the devaluation protocol. Several other value guided behavioral paradigms like Pavlovian to instrumental transfer; delayed discounting, reversal learning and other second order behaviors show a similar role for OFC (Chudasama and Robbins, 2003; Hutcheson and Everitt, 2003; Mobini et al., 2002). It is to be noted that in rodents the OFC neurons also encode for the action response that leads to a favored outcome unlike the primate OFC that only codes for the value of the outcome. Interestingly disruption of the above behaviors is also seen when lesions to the OFC are made after the animals have

(22)

22

learned the underlying cue–reward and response–reward associations. This suggests that the OFC signals are crucial for updating associative representations in other brain areas in the face of unexpected outcomes by contributing necessary information to calculate the teaching signals required to drive such learning (Niv and Schoenbaum, 2008; Wallis, 2011).

1.3.3 Ventromedial PFC: PreL and IL

The PreL and the IL are known to exert distinct opposing influences across several behavioral domains. Most studies on the role of these two areas have been conducted using fear learning and its extinction. For example, PreL lesions or genetic inactivation interferes with conditioned fear expression (Corcoran and Quirk, 2007; Laurent and Westbrook, 2009; Sierra-Mercado et al., 2011; Sangha et al., 2014). In parallel, inhibiting PreL interneurons disinhibit its output to BLA and in turn enhances fear expression. On the other hand, IL inactivation or lesions impair initial acquisition of fear extinction learning as well as its retention (Laurent and Westbrook, 2009).

Conversely, optogenetic stimulation of IL during extinction training accelerates the extinction learning and enhances its retention in subsequent retrieval tests (Do-Monte at al., 2015). Also, retention of extinction memory is dependent on gene transcription and protein synthesis in the IL (Santini et al., 2004; Mueller et al., 2008). Taken together this suggests that the PreL is required for fear expression while the IL is required for its extinction. Interestingly, recent experiments done with trace fear conditioning show that optogenetic silencing of PreL during the trace periods of the training protocol disrupts fear expression on retrieval (Gilmartin et al., 2013). It has also been reported that IL inactivation immediately following extinction training failed to prevent extinction retention (Sierra-Mercado et al., 2011). This suggests that the PreL and the IL need to be ‘online’ during fear learning and extinction respectively to mitigate their roles in flexible expression of fear memories. This is further supported by counter-conditioning paradigms where conditioned fear can be reduced by co presentation of a separate conditioned stimulus (CS) not paired with an aversive unconditioned stimulus (US) but instead paired with a positive reinforcement like a reward. In such cases inactivation of the IL (and not PreL) blocks the ability of the reward related CS to reduce conditioned freezing during retrieval.

(23)

23

The opposing roles of PreL and IL can also be extended to reward reinforced behavior.

In conditioned place preference tasks (CPP) a reward like food or cocaine is paired continuously with a spatial location in a context. Upon retrieval and in absence of the reward animals chose to spend more time in the rewarded space. In such tasks, inactivation of the PreL can attenuate CPP while stimulating the PreL enhances CPP (Moorman et al., 2014). Moreover, direct electrical stimulation of the PreL alone during training (without reward) can induce CPP while inactivation of the PreL prevents reinstatement of cocaine CPP after its extinction. Conversely, activating IL after cocaine CPP enhances its extinction while silencing it interferes with extinction (van den Oever et al., 2013). Furthermore, IL stimulation post extinction attenuates the reinstatement of cocaine CPP while its inactivation exaggerates CPP reinstatement (LaLumiere et al., 2012). Simultaneous blocking of the PreL during reinstatement can also prevent this exaggerated reinstatement of cocaine CPP. This also suggests that behavioral flexibility as shown in these tasks is dependent on the concerted action of the PreL and IL. The PreL-Go and IL-stop operational logic have also been extended to goal directed behavior where the PreL and IL are key regulators in switching between goal-directed actions and habit based responses, where the PreL is (Barker et al., 2014; Baleine and O’Doherty, 2010) required for goal-directed behavior and the IL for habit learning. This PreL-go/IL-stop property of the vmPFC creates an optimal framework to encode behavioral flexibility, which allows an animal to choose between multiple alternative response strategies based on evidence, contingencies and internal goals.

1.4. Structural plasticity in the mPFC

The development and maturation of a cortical area reflects more than a simple temporal unfolding of a genetic blueprint. Rather, it represents a complex interplay of experiential and genetic factors that mold an emerging brain area. Owing to the complexity of building a brain, which has to adapt to a specific ecological niche post birth, there is an early life overproduction of neurons and their connections. These are later sculpted by experience dependent neural activity. Thus, it is possible to use a minimum of genetic instructions to build brains that are appropriate for the specific ecological niche of an animal. In humans, the peak synaptic density in the sensory cortices is reached by the first-year post birth while the prefrontal areas reach their

(24)

24

peak by 5 years of age (Petanjek et al., 2011). Interestingly this overproduction of synapses is the highest in in the PFC among all cortical areas. Starting from early childhood these synapses undergo extensive pruning and it continues in the PFC up to the third decade of life. Rodents also share a similar trajectory of maturation of the PFC as seen in human beings (Petanjek et al., 2011). In rats, the PFC cytoarchitectonics characteristics stabilizes last at 30 days post birth when compared to sensorimotor areas, which stabilize at 24 days (van Eden et al., 1990). Thus, among all cortical areas the PFC has the strongest potential to be shaped as a consequence of childhood and adolescent experiences like social and psychological stress, sensorimotor stimulation/deprivation and hormonal changes. This in turn can radically alter the functional circuitry of the PFC and resulting cognitive functions.

1.4.1 Sensorimotor stimulation

One of the most prevalent method of studying structural plasticity is by measuring changes in dendritic arborization, spine densities, cortical thickness and expression of neurotransmitters as a consequence of housing animals in complex environments with toys, tunnels, nests, etc. over prolonged periods. In adult animals, this type of environmental complexity, also known as environmental enrichment, leads to an increase in dendritic complexity and spine densities across most cortical and subcortical structures. Surprisingly a prolonged exposure to environmental enrichment leads to no change in PFC structural plasticity in adult rats and instead a decrease in dendritic complexity in male rats (Kolb et al., 2003, Comeau et al., 2010). In a different set of experiments, early life sensory stimulation in rats improved motor and cognitive functions as well as increased dendritic complexity and spine densities in both mPFC and OFC (Richards et al., 2012). Taken together these experiments show that the maturing prefrontal is more plastic than in adulthood unlike other neocortical regions, the hippocampus (HPC), and striatum.

1.4.2 Psychoactive drugs

Some of the largest and most robust changes in prefrontal cortex structural plasticity are caused by repeated administration of psychoactive drugs. When animals are given repeated doses of psychomotor stimulants, there is strong behavioral effect

(25)

25

(hyperactivity) that is accompanied with dendritic changes in prefrontal cortex and NAc (Robinson and Kolb, 1999a,b, 2004). For example, when animals got repeated low doses of amphetamine there was an increase in dendritic length and spine density in mPFC and NAc but a corresponding decrease in these measures in OFC (Crombag et al., 2005). Interestingly the differential effect of psychoactive drugs on the mPFC vs OFC also hold true for other psychoactive substances such as cocaine, and tetrahydrocannabinol. These changes are not restricted to pyramidal neurons alone, as one study found that a single injection of methamphetamine in adulthood led to a 20% increase in the density of GABAergic synapses in the PFC (Dawirs et al., 1997).

The above mentioned anatomical changes are also accompanied by changes in gene expression in the PFC. Specifically amphetamine injections increased the expression of Fibroblast growth factor-2 (FGF-2) in the PFC (Cuppini et al., 2009). This factor is proposed to reduce neuronal excitability by inhibiting voltage gated sodium and potassium currents. Thus, the PFC shows rapid and persistent changes in their ultrastructure and gene expression in response to psychomotor stimulants.

1.4.3 Stress

Stressful experiences have been a second focus of recent studies in structural plasticity in the PFC. For example, in male rats chronic stress dramatically reduces synapses in layer III pyramidal neurons throughout the mPFC (Cook and Wellman, 2004; Radley et al., 2006, 2008; Bloss et al., 2010, 2011). Interestingly the synaptic loss in young animals can recover within 3 weeks post the stressful experience but older animals fail to show this recovery (Goldwater et al., 2009). Along with chronic stress, short-term mild stress can also result in structural changes in the PFC like retraction of dendrites. It is interesting note that the PFC ultrastructure is inherently different in males and females due to sexual dimorphism. Accordingly, the effect of stress on structural plasticity in the PFC is different across the genders (Muhammad et al., 2012). A recent study on the epigenetic effects of adult stress in mPFC, OFC, and the hippocampus showed increase in global methylation in both sexes in prefrontal areas but a decrease in the hippocampus. When examining the RNA from the above brain regions, chronic stress exposure led to mostly non-overlapping changes in gene expression across the different sexes and brain regions. These epigenetic data are consistent with the morphological data in showing both regional

(26)

26

and sex differences in the effect of stress on the prefrontal cortex plasticity (Kolb et al., 2014). From these two critical points emerge about the prefrontal cortex. First, unlike other cortical areas, the PFC shows age dependent changes in structural plasticity in response to sensorimotor stimulation, where young and adolescent animals have a more plastic PFC. Second, owing to the prolonged maturation timeline of the PFC post birth, experience dependent plasticity can strongly change their neurodevelopmental trajectory and subsequent function in adult hood.

1.5. Memory processes within the mPFC

1.5.1 Synaptic rearrangements in memory formation

The neural mechanisms for memory formation and storage can be classified into two related theories: the ‘synaptic plasticity theory’ postulated by Donald Hebb and ‘the cellular engram theory’ proposed by Richard Semon (Semon 1904, 1909). The very first proponent of the synaptic theory of memory formation was Santiago Ramon y Cajal, who suggested that contacts between neurons as site of memory storage (Ramon y Cajal, 1893). Subsequent work by Donald Hebb suggested a theoretical mechanism where synapses between neurons are strengthened by co-ordinated activity by the participating neurons (Hebb, 1949), now referred to as spike-time- dependent plasticity. Seminal work done by Eric Kandel using the Aplysia siphon- withdrawal reflex provided the first evidence for the synaptic theory of learning (Kandel et al., 2014). Subsequent discovery of long-term synaptic potentiation in the 1970s elucidated the molecular events that lead to synaptic strengthening on coordinated activity (Bliss and Lomo, 1973). The most commonly studied form of LTP is NMDA receptor dependent LTP where a large influx of calcium ions into the post synapse through the NMDA receptor leads to clustering of AMPA receptors and a resulting enhancement of synaptic transmission. Disrupting part of this process, for example by blocking AMPA receptor trafficking, leads to disruption of memory formation (Nabavi et al, 2014; Kessels et al., 2009). In addition to synaptic strengthening learning also leads to structural changes such as increased spine turnover and synaptic rearrangements (Caroni et al., 2012, 2014).

(27)

27 1.5.2 Cellular basis of memory allocation

The idea for a ‘connected complex of excitations that exist as a unified engram complex’, as the basis for a memory trace, was first proposed by Richard Semon in the early 1900s (Semon 1904, 1909). This theory envisions a system whereby a part of the original learning condition (Eg: a cue) can reactivate the engram complex and thus provide a framework for memory storage and retrieval (Schacter et al., 1978).

Studies over the last 40 years have provided us with strong evidences towards a cellular basis for memory engrams, where subpopulations of neurons in a brain area or multiple brain areas store individual episodic memories. A notable work from the 1980s used single cell activity in the inferotemporal cortex of monkeys performing a visual delayed matching-to-sample task. The authors demonstrated that groups of neurons differentially responded to color depending on attentional processes and this activity was correlated to retention and retrieval of the visual memory (Fuster and Jervey, 1981). Further evidence for a cellular basis of memory formation came from examining the expression of immediate early genes (IEGs) like c-fos and Zif268 where populations of cells active during the initial learning paradigm were selectively reactivated during memory retrieval (Reijmers et al., 2007, Tayler et al., 2013, Zelikowsky et al., 2014). These observational studies have been recently corroborated by gain of function and loss of function studies. By overexpressing the transcription factor CREB in a random population of amygdalar neurons immediately prior to learning a group of researchers could bias the storage of the corresponding memory to these neurons. Subsequent ablation of these neurons led to a loss of the encoded memory as seen by failure to retrieve the memory (Han et al, 2007, Han et al., 2009, Zhou et. al., 2009). Finally, gain of function studies have further strengthened the cellular theory of memory storage. In a recent study, researchers were able to genetically tag a population of neurons active during learning and subsequently reactivate them in the absence of any original learning cues to result in behavioral retrieval of the memory (Liu et al., 2012, Kim et al., 2014; Yiu et al., 2014). Thus, the synaptic plasticity theory and the cellular engram theory together provide a conceptual framework for encoding, storage and retrieval of memories.

(28)

28 1.5.3. Memory consolidation

A newly formed memory might exist for minutes or days or years before it is retrieved.

To support such a varied degree of persistence, several neurological processes have evolved which operate at all levels, namely, neuronal networks, structural plasticity and molecular levels to form a stable memory. Memory consolidation is an umbrella term describing these processes.

Memories can be broadly divided into short-term (minutes,) and long-term (days, years) memories. Short-term memories do not depend on transcription and synthesis of new proteins, while de novo protein synthesis is required for long-term memories (Bekinschtein et al., 2007). The process of long term consolidation of memories is triggered by the initial learning itself and involves early (minutes) and late (hours) events to form stable memories by strengthening and formation of synapses (Caroni et al., 2014; De Roo et al., 2008; Holtmaat & Svoboda, 2009; Ruediger et al., 2011;

Takeuchi et. al., 2013). In the early stages of memory consolidation, a learning event triggers neuronal depolarization and an inﬂux of Ca2+, which initiates a downstream molecular cascade that results in transcription and translation of plasticity-related proteins (PRPs). In turn, these PRPs induce structural and functional changes in local neuronal networks, resulting in new, remodeled, or strengthened synaptic connections among an assembly of neurons, which forms the memory engram.

Among the PRPs, the immediate early genes (IEGs) are of special note. IEGs such as c-Fos, Arc, Zif268 have been linked to long-term consolidation (Katche et al., 2010 and 2013; Nakayama et. al., 2015). As shown for c-Fos, the IEG transcripts are expressed within minutes after learning while the protein products can be detected forty-five minutes after the initial learning event and remain upregulated for up to 4h (Karunakaran et. al., 2016). Since most IEGs are transcription factors (c-Fos, Zif268) or actin cytoskeleton remodelers (Arc), they likely play a role in formation and strengthening of new synapses through epigenetic mechanisms, differential gene expression and de novo synthesis of synaptic proteins e.g. scaffolding proteins of the post synaptic density (Holtmaat & Caroni, 2016). Currently it remains unclear whether the entire cell population active (by IEG expression) during the original memory acquisition event are reactivated on memory retrieval.

(29)

29

Interestingly a second wave of IEG and other signaling factor expression is seen hours post-acquisition in a late window of memory consolidation. For example, phosphorylated MAP kinase levels, which peak around 15 minutes post memory acquisition also shows a second peak at 9h post-acquisition (Trifilieff et al., 2006).

Similarly, c-fos is also detected in a second wave of IEG expression 12-15h post memory acquisition. (Katche et. al., 2010). While the precise role of this second delayed window of consolidation remains unclear it is functionally essential for long- term persistence of memories. Recent work has shown that interfering with local dopaminergic signaling or de novo protein synthesis in this window in the mPFC or the hippocampus can disrupt later retrieval of aversive memories (Rossato et. al., 2009; Gonzales, 2014). Interestingly this deficit in memory retrieval can be reverted by BDNF administered in this same window (Bekinschtein et. al., 2007). Recent work has also shown that learning induced plastic changes in Parvalbumin interneurons are crucial in this window for long-term persistence of memories (Girardeau et al., 2009;

Karunakaran et al., 2016). Other cellular and synaptic proteins are also upregulated in this delayed window albeit with different temporal scales. For example, a second protracted peak of Arc expression is seen in the hippocampus CA1 area 8-24h after spatial exploration (Ramirez-Amaya et. al., 2005). In conclusion, this second late and protracted window of memory consolidation involves a variety of transcriptional, translational, signaling and structural changes, which are essential for persistence of memories and have been suggested to support flexible learning. (Katche et al., 2013)

1.5.4. mPFC in short-term memory

Numerous studies have implicated the mPFC, like the primate dorsolateral PFC (dlPFC) in a form of short-term memory, which spans over minutes and is referred to as working memory. Similar to primates with damage to dlPFC, rodents with mPFC damage also show deficits in tasks requiring a delayed response (Horst and Laubach, 2009). The role in working memory is further supported by observations that both, rodent mPFC and primate dlPFC, exhibit persistent cellular activity during delay periods that is selective for a prior or upcoming target location (Baeg et al., 2003;

Funahashi, 2006). Another study showed that when rats were required to hold down a lever until cued; a third of mPFC cells significantly altered their activity during the delay (Narayanan and Laubach, 2006). However, in general the mPFC is not required

(30)

30

to form short-term memory for locations, odors, or objects (Birrell and Brown, 2000;

Seamans et al., 1995). It is interesting to note that short-term memory for rewarded odors do depend on mPFC when either a large number of odors must be remembered or odor associations must be learnt through social interaction (Boix-Trelis et al., 2007).

Thus, the mPFC is only required for a very specific form of short-term memory called working memory or if the demand for short-term memory storage is very high.

1.5.5. mPFC in long-term memory

Early imaging studies first indicated a specific role for mPFC in long-term memory. A study examining metabolic activity in multiple mouse brain regions during retrieval of rewards on an eight-arm maze, either 5 or 25 days after learning, showed that the mPFC had significantly more activity during remote retrieval compared with recent.

This selective activation of mPFC in remote memory has also been replicated in tests of both spatial and fear memory (Frankland et al., 2004; Teixeira et al., 2006).

Consistent with these imaging results, inactivating mPFC also leads to deficits in retrieval of remote memories while leaving recent memory intact (Frankland et al., 2004; Ding et. al, 2008). However, several lines of evidence support the involvement of mPFC in recent memory. Of note is a recent study where optogenetic silencing of the PreL during the trace period of trace fear condition led to disruption of memory retrieval the next day (Gilmartin et al., 2013). Other studies have also directly demonstrated the requirement of mPFC for retrieval of recent navigational (Churchwell et al., 2010), object-place (Lee and Solivan, 2008) and fear memories (Corcoran and Quirk, 2007), learned a day or two prior to testing. It is not clearly understood why in some experiments the mPFC is not required for recent retrieval. However, an explanation for this could be that mPFC serves for both storage and representation of a memory on different time scales. During retrieval of recent memories, the role of mPFC is to represent context, events and responses while the mapping between them is stored within the hippocampus. On the other hand, during remote retrieval, the mPFC does both, i.e. represents as well as stores context-event-response mappings.

Thus, the brain may be less able to compensate for its loss during remote retrieval than during recent. (Euston et al., 2012)

(31)

31 1.6. Oscillations and mPFC function

The brain uses electromagnetic oscillations as a way of linking ongoing processing across multiple brain areas. To this end, a wide variety of frequencies are involved, from the slow oscillations seen during sleep (starting at a fraction of a Hertz) to high frequency gamma oscillations reaching 80–100 Hz. By transiently increasing and decreasing the degree by which oscillations in the two structures keep a constant phase relationship, different areas can regulate the extent to which their computations interact. This allows for a dynamic functional network whose participating members can change rapidly along with task demands. For example, by shifting the phase relationship between gamma oscillations, a higher visual area like V4 can tune itself on a V1 column whose receptive field contains a preferred stimulus, thereby steering visual attention (Womelsdorf et al., 2007) to that stimuli. Furthermore, oscillations at different frequencies may coordinate incoming information from different sources. For example, the hippocampal subfield CA1, can selectively tune in to signals from the CA3 subfield or the entorhinal cortex by oscillating at characteristically different gamma frequencies inherent to these structures (Colgin et al., 2009).

Among the different types of oscillations, gamma oscillations have been strongly linked to the working memory function of mPFC. Specifically, gamma oscillations reflect rhythmic firing of inhibitory interneurons, particularly parvalbumin (PV)- expressing fast-spiking interneurons. The synchronized rhythmic patterns of spiking and synaptic inhibition of these PV neurons give rise to the electrical signatures identified as gamma oscillations. Interestingly, the intrinsic and synaptic properties of fast-spiking PV interneurons seem to be tuned to produce gamma oscillations at specific frequencies (Galarreta and Hestrin, 1999; Gibson et. al., 1999; Bartos et al., 2002). Mechanisms through which gamma oscillations contribute to circuit function can be classified based on locus of their action. Gamma oscillations can act on local excitatory neurons to regulate their responses to incoming stimuli, to modulate the patterns of local circuit activity or to enhance the efficacy of their output. In parallel, gamma oscillations can act as a clock relative to which the precise timing of spikes of distributed neurons can carry information and indicate their participation in a common neural representation (Fries et. al, 2007; Singer 1993).

(32)

32

In the PFC gamma oscillations increase during tasks that involve attention (Kim et al., 2016), cognitive control and social interaction (Cho et al., 2015). Moreover, the power of these oscillations also scale up with task demand. For example, the power of gamma oscillations in the PFC increases with the number of items being stored (Roux et al., 2012). Furthermore, a recent study showed that optogenetic stimulation of mPFC PV interneurons at 30–40 Hz improved performance, whereas stimulation at 1–10 Hz produced a much larger decrement in performance in an attention task.

Although these oscillations are generated locally, they can be synchronized across long distances and correlate well with behavior. For example, synchronization between spiking activity in PFC neurons and hippocampal gamma oscillations is necessary for successful encoding of spatial information in a working memory task (Spellman et al., 2015; Tamura et al., 2016; Sohal, 2016).

1.7. The Hippocampal-Prefrontal axis

Several direct and indirect pathways link the hippocampus and the PFC. In both rodents and primates, the PFC receives monosynaptic glutamatergic connections from the hippocampus (Condé et al., 1995; Hoover and Vertes, 2007, Thierry, 2010).

These originate only in the ventral hippocampus (vH) and target the mPFC, with stronger projections to ventral sub regions (Condé et al., 1995; Hoover and Vertes, 2007). Recently a monosynaptic connection originating from the anterior cingulate cortex and terminating in CA1 and CA3 of dorsal hippocampus (dH) has also been identified (Rajasethupathy et. al., 2015). Additionally the hippocampus and PFC can also interact via indirect routes, like through the thalamic nucleus reuniens (NR) which reciprocally connects to vH and mPFC (Vertes, 2006; Cassel et al., 2013). A second indirect rout of communication could be through the lateral entorhinal cortex, which also connects reciprocally to both the structures (Moser et. al., 2010).

(33)

33

Figure 1.5: The hippocampal-prefrontal axis:The vH sends monosynaptic projections to the vmPFC while the ACC projects to the dH directly. The vH-vmPFC connection has been implicated in expression of learnt fear and working memory while the ACC-dH projection is involved in fear memory retrieval.

The hippocampus and prefrontal cortex can also communicate via the reuniens and lateral entorhinal cortex. (Sigurdsson and Duvarci, 2016)

Hippocampal and prefrontal neurons often spike within a short time (~100 ms) of each other (Siapas et al., 2005). Furthermore, these spikes in prefrontal neurons can lead or lag behind spikes in the hippocampus, showing a directionality of influence (Siapas et al., 2005; Wierzynski et al., 2009). Prefrontal neurons are also modulated by the phase of hippocampal theta oscillations (4-12Hz) (Buzsáki, 2002). Prefrontal neurons tend to fire more at certain phases of the theta oscillation, a phenomenon referred to as “phase locking” (Siapas et al., 2005; Sigurdsson et al., 2010). Such phase-locked PFC neurons also show cross-correlations with hippocampal neurons, suggesting they reflect the same phenomenon (Siapas et al., 2005). Interestingly, prefrontal neurons can be phase-locked more strongly to either past or future phases of hippocampal theta oscillations although on average phase-locking to the past is stronger (Siapas et al., 2005; Sigurdsson et al., 2010). Lastly, hippocampal-prefrontal synchrony can be observed in the “coherence” of LFPs recorded in the two structures (Adhikari et al., 2010).

Circuit mechanisms for learning in the rodent Prefrontal cortex and their dysfunction in Schizophrenia